Low-latency prediction of critical values of currents and electromagnetic torque in synchronous generators under line-to-ground short circuit by data sparsification

In the topical context of real-time contingency analysis, which is to be carried out in an energy control centre, and requiring fast evaluation of thousands of scenarios, storing the full trajectories for all scenarios overwhelms databases. Moreover, during electrical transients, control actions are triggered based on the detected critical values of the various characteristic quantities of the processes involved. With these in perspective, the approach in the present paper enables the real-time transient analysis of the highly representative scenario of line-to-ground short circuit fault in synchronous generators. Since real-time digital simulation does not allow of post-processing latency, instead of uniform time-step recording followed by post-hoc critical values detection, we make use of the fact that the state variables critical values occur precisely where the rates of change of state variables, encompassed by a state-space model, cross zero. The extension to any non-state quantity, like the generator electromagnetic torque, by computing its rates of change on-the-fly appears as a natural generalization.